Paper 2018/1123
When Theory Meets Practice: A Framework for Robust Profiled Side-channel Analysis
Stjepan Picek and Annelie Heuser and Cesare Alippi and Francesco Regazzoni
Abstract
Profiled side-channel attacks are the most powerful attacks and they consist of two steps. The adversary first builds a leakage model, using a device similar to the target one, then it exploits this leakage model to extract the secret information from the victim's device. These attacks can be seen as a classification problem, where the adversary needs to decide to what class (corresponding to the secret key) the traces collected from the victim's devices belong to. For a number of years, the research community studied profiled attacks and proposed numerous improvements. Despite a large number of empirical works, a framework with strong theoretical foundations to address profiled side-channel attacks is still missing. In this paper, we propose a framework capable of modeling and evaluating all profiled analysis attacks. This framework is based on the expectation estimation problem that has strong theoretical foundations. Next, we quantify the effects of perturbations injected at different points in our framework through robustness analysis where the perturbations represent sources of uncertainty associated with measurements, non-optimal classifiers, and methods. Finally, we experimentally validate our framework using publicly available traces, different classifiers, and performance metrics.
Metadata
- Available format(s)
- Publication info
- Preprint. MINOR revision.
- Keywords
- Machine LearningRobustness AnalysisSupervised LearningFramework
- Contact author(s)
- picek stjepan @ gmail com
- History
- 2021-06-09: last of 4 revisions
- 2018-11-20: received
- See all versions
- Short URL
- https://ia.cr/2018/1123
- License
-
CC BY